CHW entrevista a Jen-Hsun Huang
Our tech media was invited to interview NVIDIA’s founder and CEO, Mr. Jen Hsun Huang. It happened a few days ago during the GPU Technology Conference event, formerly known as NVISION.
The following text is the transcription of a 30 min meeting with JHH. We’ve tried to preserve the spirit of the conversation but we also took out most trivial comments. Mr JHH answered each and every question spontaneusly and that’s why some of them seem more like a reflection instead of a statement.
CHW: What will drive NVIDIA’s growth in the future? Hardware? Software? GeForce? Tegra? Tesla? Etc…
JHH: Well, NVIDIA is a software company, this will be what will drive our growth, we just don’t write de application, but we create all the core technology we create more core technology for visual computing that any other company in the world. We give it away, it’s like Google they do search but they give it away, they sell advertisement.
What you are and how you make money doesn’t have to be the same. NVIDIA is a integrated complete visual computing and parallel computing solution technology company.
CHW: How do you make money then?
JHH: By selling hardware, you do know that Apple is a software company but they make money by selling hardware. In the future software is the most important thing, anyone can make chips, it is really expensive but anyone can do it. The hard part is to inspire people to create amazing things.
We have software, systems, architecture, tools, compilers, and languages, whatever it takes…
CHW: How do you protect your market share?
They way that we protect our market share or another way of thinking about it is how we become exceptional in our market share is to offer more, so when you have Tegra your performance is better, even though it is the same chip manufactured at TSMC our chip is better because the architecture is better.
Other companies like TI, Qualcomm, Freescale they would license pieces of technology from ARM and other people because we have the ability to built it ourselves our solution is more integrated it works better together.
CHW: How do you plan to always be the best? How do you achieve this?
JHH: The best way to do it is to have no backup plan, let me give you an example, when you are flying trapeze, swinging… they say that you should pretend that there is no net, now of course when they fall down is good to have a net, for us we have no net, my employees know there is no net, there is no safety, there is no back up, as a result you have to create good things that is how we became NVIDIA.
Now we are not always the best, but we always try to be the best the absolute best, over the course of the last 15 years on average we have been absolutely the best, what continues to challenge ourselves is to be more creative more imaginative. There are no warranties but that is why you need to have such good people.
CHW: It’s been 10 years since you introduced the GeForce brand, that is when the GPU started to replace the CPU in some areas, how would you describe the world of computing then, now and how do you imagine it in the future.
JHH: 10 years ago when we launched the GeForce brand the vast majority of the content in the application used the CPU, the GPU only for graphics acceleration, so triangle set up, triangle transform, lighting very simple lighting, very simple shadows, texture mapping, multi texture those specific functionality was put into the graphic pipeline, the first pipeline that people described was the OpenGL pipeline and then Microsoft developed the DirectX pipeline, this pipeline allowed the graphic chips to do specific thing really specific, the graphics chips was not very smart it could do only one thing but it was so fast, so good and so well that it became so popular.
In order for a CPU to replace the GPU it should be a 1000 times faster, too much way to much and that is why the graphics market grew and grew and grew even though the CPU keep getting faster, the GPU was so fast it was impossible to catch up and then of course game developers discovered how wonderful is for graphics to be instantaneously fast so they made more and more beautiful graphics, if you look at Quake 3 and Batman today you can’t even compare it is impossible, right? So the content developers fell in love with basically an infinitely faster graphics subsystem, so the content kept coming and becme more and more beautiful, so the content increased, the resolution increased, people wanted more Anti Aliasing then the GPU requirement kept going up and up.
About 7 years ago I started to see that if you make the graphics chip no matter how fast but it does one thing (texture mapping) eventually every content developer in the world would have the same picture or look the same but very very fast, if you look at older texture map games the images is really flat, every single game looked the same there was no style no art, so we decided to make the GPU more programmable and that was the introduction of the programmable shader but we added just a little bit of programmability with just this small amount of programmability game developers went crazy because their game looked different, completely different and now programmability came into the world, what it ultimately did was to add artist expression to computer graphics, so we brought style, we brought art to the computing gaming industry.
Once you made the art form artistic they want infinite abilities to tell a story, at this point they said we want all to be programmable. They don’t want to see hundreds of shaders they want to have hundreds of processors, so they can write any program they want. In 10 years time every single CPU will be heterogeneous, every single CPU will have a few CPU cores and GPU cores, no question about it. The question is only are they all in one chip? Absolutely no sometimes you have it in one chip, some in 2 chips, sometimes 1 CPU and many GPUs and in a server the mix is different, it depends on the need, this is the future.
I believe this architecture that we made will last the industry 30 years, because the last architecture lasted 30 years. Once content developers start to learn an architecture they don’t want to change it because you want that the application you write today work in the future.
CHW: What are the plans for the Chipset business unit?
JHH: Our chipset business unit is going to introduce some very exciting technology in the future, next year is going to be a grow up year for this unit but because of Intel claims in the market place, it’s very disruptive tactics in the market it is impossible for us to be successful building chipsets with DMI for the next generation CPUs, so we will exit the market, our chipset unit business will continue to work for well over 1.5 or 2 years, we will introduce some interesting new technology that is not a Chipset or a GPU so we are very excited about that, for enhancing Atoms and enhancing the current generation of CPUs, we are excited about this but we haven’t announced it yet.
We are going to support the Pentium 4 Front side bus for as long as they are in the market place.
CHW: Are Gamers the most profitable market for NVIDIA?
JHH: I would say probably not, 2/3 of our revenue come from games but 2/3 of our profit comes from Quadro workstations, you know NVIDIA is the words largest workstation technology provider, our market share is about 95% in this market, every single company, every single country, every single car, oil, plane, train, shoes, etc… uses Quadro technology, this is the highest market share of any workstation company in the history of making but the workstation is not very large but is very important.
CHW: There is a lot of expectation from Tegra, when can we expect Terga based Smartphone’s?.
JHH: We have many Smartphone’s been built, they will be out in Q1, Q2 and Q3 every quarter you’ll see new Smartphone’s based on Tegra.
CHW: Who are you working with to make Smartphone’s, any big players? Because we’ve seen a couple of Chinese OEMs working on Smartphone’s based on Tegra. We know that Apple has hired a lot of people to develop chips and Nokia has a partnership with Intel to make Atom based Smartphone’s.
JHH: I can’t announce them but they are big players. There are a lot opportunities outside of Apple, the iPhone is a really successful product but there are a lot of opportunities outside of Apple.
Every single Phone in the future will become a Smartphone, the market will be huge, we will be ready for this market.
CHW: Are you going to license Tegra or parts of it? Like ARM does.
JHH: No we have no reason to license it, if you really like GeForce just buy Tegra it is easy to buy. It is so small.
CHW: What do you expect from ARM, in CPUs and GPUs? Do you see them as competitors?
JHH: We are friends with them, Tegra has a GPU core inside and has a ARM core inside but that is not what Tegra is, it’s processor it has 8 cores inside, it is core processing at the extreme but what people buy is not a ARM CPU or a GPU, they buy the experience that Tegra gives.
ARM licenses their technology to my competitors but they don’t compete with me.
CHW: Been a gaming company was it difficult to get the attention from the scientific community?
JHH: First of all, we are associated to a gaming company because GeForce as a brand is so strong and when you do something very very well for a long time and you do it with a great deal of passion for the people that you care about they invite you to their home.
But we are not actually a gaming company we are a video computing company, but we have cared so long for gamers that they invited us to our homes, they are one of us.
Of course there is challenge that the scientific community wouldn’t take us seriously but don’t forget that many of my customers are Scientists. Yesterday you saw a presentation from the CEO of Technyscan, this company does breast cancer early detection, the way they found out about CUDA was from the engineer and he found out about CUDA was because he was a gamer. So when we invented CUDA instantaneously he understood how he can help people with CUDA, it took him almost one year to convince the company to use CUDA, he was our best sales men, remember there are millions and millions of gamers out there and they are smart, so you want the gamers from your side.
CHW: What do you think about the Lucid Hydra 200 Chip?
JHH: I don’t know, it sounds like terrible idea. There is so much software going on, drivers, applications and optimization that to think that there is no software is a bad idea, too much to expect. Plus SLI and Crossfire are available to anybody.
CHW: Can we talk about the next generation? I get the impression that Fermi won’t be as good for gaming as it will be for parallel computing, can you comment on that.
JHH: You don’t have to worry, we presented Fermie this as a parallel processor, way because GTC is not a gaming event, it is a parallel computing event.
The new graphics capabilities of the new GeForce has not been revealed yet, we have some really exciting technologies to share, of course one of the most interesting technologies we shared it yesterday, in the future there will be rasterization and raytracking you won’t have to choose.
There are also a lot of other exiting secrets that we will talk about later.
CHW: Will Fermi be ready before for gamers?
JHH: Yes, testing for gamers take the same amount of time that for scientists, there is nothing more frustrating that wanting to game and the computer crushes. The time to get the hardware to market is really fast for gamers.
We want to thank NVIDIA and Mr Jen-Hsun Huang for their time and kindness.